Labelbox

Labelbox is a web and API platform for AI teams, AI labs, and enterprises that need a data factory for curating datasets, labeling multimodal data, and running evaluations for model training and post-training.

Key capabilities include:

• Catalog for data management and curation (usage tracked in Labelbox units (LBUs), including 1 LBU per 60 data rows)
• Annotate for multimodal annotation and labeling, with quality controls for consistent outputs
• Model-assisted labeling and AI predictions to accelerate labeling workflows (including LBU-based usage such as 1 LBU per 5 data rows or per labeled row)
• Custom evaluations for tasks ranging from reinforcement learning data to private benchmarks and head-to-head comparisons

Pricing: Freemium with a Free tier; consumption-based monthly billing on LBUs (Starter rate about $0.10/LBU with volume discounts; Enterprise is negotiated). Managed labeling services are available, including Standard, Alignerr, and Alignerr Connect, starting at $10/hour.

Notable for supporting end-to-end workflows, from dataset curation and model-assisted labeling to RLHF/SFT-style post-training tasks, plus access to the Alignerr expert network for complex human-in-the-loop data generation and red teaming.